BSite-pro: A Novel Approach for Binding Site Prediction in Protein Sequences

Accurate protein binding site annotations are vital for a profound understanding of biological processes and protein interactions. Due to less supporting information a large number of proteins remain uncharacterised. For large sets of unchar- acterised proteins, only amino acid information is available. In this paper, we proposed BSite-pro - a traditional approach - which makes use of protein sequence data for classification of binding sites. The classification procedure requires hand-crafted features from protein sequences. Our results stipulate noteworthy enhancements concerning predicted accuracy and recall upon comparison with previously proposed sequence-based techniques. BSite-pro achieves an overall validation accuracy of 85.06% and recall of 82.17%. Finally, we discuss that using the same feature extraction methods and model, we profitably dealt with two contrasting types of problem i.e. protein active and conserved sites prediction.